Mediary

Inspiration

Mediary was inspired by a common HR problem: workload overload is often invisible until it becomes burnout.

In many teams, the warning signs are scattered across different signals such as meeting load, after-hours work, uneven task distribution, and employee self-assessments. Individually, these signals may look normal. Together, they can reveal a pattern of sustained overload.

Mediary is an AI workplace assistant that detects invisible overload patterns before they become burnout.

What it does

Mediary analyzes workload signals across a simulated organization, identifies employees or teams under strain, explains the strongest risk drivers, and recommends the right follow-up action.

The demo includes two scenarios:

• Normal workload scan, where workload pressure is manageable.
• High overload simulation, where sustained strain is detected and routed into an HR intervention queue.

The goal is to help HR teams move earlier, before workload issues become serious.

How we built it

We built Mediary as a deterministic agent workflow inside a Next.js dashboard.

The system follows three main roles:

1. A signal reader that collects workload inputs.
2. A risk analyzer that scores workload strain.
3. A workflow coordinator that recommends the next action.

The dashboard shows the agent execution trace, risk distribution, individual workload signals, and recommended HR follow-ups. We also added a streaming endpoint so the demo can show the agent workflow progressing step by step.

Challenges we faced

The main challenge was wording. Since burnout is a sensitive topic, Mediary avoids medical claims. It is designed as an operational workload assistant for HR teams, not a diagnosis tool.

What we learned

We learned that agents are more valuable when they help users take action, not just generate insights.

For HR teams, the useful output is not only “this person may be overloaded.” It is also “why this looks risky” and “what should happen next.”

What's next

Mediary could be expanded with real integrations like Google Calendar, Slack, HRIS data, and pulse surveys. Future versions could include trend tracking, configurable escalation rules, and team-level workload forecasting.

Built With

• hermes
• mimo
• typescript